AI Industry Faces Reckoning as Costs Rise with Usage-Based Billing Models

AI companies have long lured customers with affordable—or even free—access to their models. But the era of unchecked generosity is ending, and the transition is proving costly for users. Earlier this month, Microsoft’s GitHub Copilot customers received notices about reduced usage limits due to “significant strain” on company servers. Free trials for new accounts were also eliminated, citing system abuse.

Now, GitHub is doubling down on cost-cutting measures. On Monday, the company announced that all GitHub Copilot plans will transition to usage-based billing, effective June 1. The new model charges customers based on the number of tokens consumed during AI tasks, replacing the previous “premium request units” system.

Under the old model, users paid a fixed fee for a set number of intensive “premium” requests, regardless of the actual computational cost. The new system ties pricing directly to token consumption. Subscribers receive AI credits equal to their monthly subscription fee—for example, a $10-per-month plan includes $10 in monthly AI credits. Additional credits can be purchased as needed.

In the announcement, Mario Rodriguez, GitHub’s chief product officer, called the old pricing model “no longer sustainable.” He explained,

“A quick chat question and a multi-hour autonomous coding session can cost the user the same amount. GitHub has absorbed much of the escalating inference cost behind that usage.”

The shift reveals a harsh reality: AI companies can no longer afford to subsidize heavy usage. Until recently, they absorbed the costs of providing cheap access to power-hungry models, attracting millions of users. But today’s AI tools—especially coding assistants and autonomous agents—demand far more resources. Companies are rapidly deploying these tools across workforces, encouraging employees to use them extensively. Software engineers, in turn, often run multiple AI agents simultaneously, generating code for hours while racking up backend costs.

Industry-Wide Crackdown on AI Usage

GitHub Copilot isn’t alone in tightening its policies. Anthropic has repeatedly adjusted Claude Code rate limits, recently imposing stricter caps during peak hours. The company also experimented with cutting off access for its lowest-tier paid users, sparking backlash. Earlier this year, Google introduced weekly limits on its AI coding environment, Antigravity.

As financial pressures mount, more AI companies are likely to follow suit. The question remains: How will customers and businesses adapt to a new era of transparent, usage-driven pricing?